Keynote: Kubernetes Serverless Present and Serverless Future
summary
Argues Kubernetes should hide machines so users only think about workloads and describes a working prototype: the virtual kubelet registers a fake node with the API, mediates scheduling into serverless providers and places containers onto serverless containers (providers listed: Azure Container Instances, Alibaba ECI, Hyper.sh, Amazon Fargate, VMware). Notes orchestration APIs (deployments, replica sets, cronjobs) remain unchanged while scheduling moves out, raises open questions about affinity, networking and failure domains, and announces AKS now runs in 16 regions including Azure China private preview.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 4:26Server Lists
- 5:57Kubernetes API
- 7:00Programming Models
- 8:21Serverless Infrastructure
- 9:28Orchestration vs Scheduling
- 12:24Serverless Container Infrastructure
- 15:34Open Questions
- 16:57Virtual Cubelet
- 18:53Conclusion
related talks
11mKeynote: Containerful Serverless
15mKeynote: Kubernetes
32mThe Path to a Serverless-Native Era with Kubernetes
39mServerless on Kubernetes with Kubeless [A]
29mServerless on Kubernetes: What We Learned From Customers
26mKeynote: Kubernetes - The Universal Control Plane
39mDelivering Serverless Experience on Kubernetes: Beyond Web Applications
20mKnative: A Kubernetes Framework to manage Serverless Workloads